On Fri, Mar 31, 2017 at 01:30:15PM -0700, Omar Sandoval wrote:
> On Fri, Mar 31, 2017 at 04:30:44PM -0400, Keith Busch wrote:
> > On Fri, Mar 31, 2017 at 11:59:24AM -0700, Omar Sandoval wrote:
> > > @@ -2629,11 +2639,12 @@ void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues)
> > > set->nr_hw_queues = nr_hw_queues;
> > > list_for_each_entry(q, &set->tag_list, tag_set_list) {
> > > blk_mq_realloc_hw_ctxs(set, q);
> > > - blk_mq_queue_reinit(q, cpu_online_mask);
> > > + blk_mq_queue_reinit(q, cpu_online_mask, true);
> >
> > I think you want to call blk_mq_update_queue_map directly outside this
> > loop rather than for each queue through blk_mq_queue_reinit. We only
> > need to map the queues once per tagset rather than per queue.
>
> Right, thanks, I'll do that. I figure you're the person to ask,
> nvme_add_dev() does want the remap to happen, right?
Yep, nvme may want to change the queue count if you alter either the CPU
topology or some device specific setting to reprovision hardware queues.
